With the release of January 2019 Consumer Price Index (CPI) data, the U.S. Bureau of Labor Statistics (BLS) began to adjust the collected price quotes of residential telecommunications services to ensure they are of constant quality. In estimating the CPI, BLS measures only differences in price between comparable items, not differences in price that arise from differences in quality. As part of ongoing efforts of BLS to improve the accuracy of the CPI series, this article describes the challenges that led to the adoption of methodological changes to the CPI series for residential telecommunications services and briefly discusses the updated models that incorporate them.
Defining residential telecommunications services
The residential telecommunications services group comprises the indexes for residential telephone services; internet services and electronic information providers; and cable, satellite, and live streaming television service. It is now adjusted to account for the rapid technological change in these services.
The residential telephone services series includes all types of local, long-distance, and Voice over Internet Protocol (VoIP) residential telephone services. VoIP operates similarly to other residential telephone services, except that VoIP requires an internet connection. Over time, providers of these services have tended to increase the amount of allowed calling, with many firms adding unlimited calling. Many other features have also been added to available packages.
The internet services and electronic information providers component of the CPI is composed of charges for internet access through digital subscriber lines (DSLs), cable services, and fiber-optic services, as well as other online services such as web hosting, domain names, and file hosting for noncommercial use. Wireless telephone plans that include internet access are not eligible in this category (such plans are included in the wireless telephone service component). However, mobile internet access provided by wireless carriers through equipment, such as mobile hotspots, is eligible to be included. These plans can change rapidly with respect to upload and download speeds.
The CPI series for cable, satellite, and live streaming television service contain subscription fees for services such as basic cable, digital cable, expanded cable, streaming video services provided by cable or satellite television providers, live television on internet-connected devices, and premium movie channels. The category does not include third-party subscription video services or satellite radio services. Different television service packages can vary widely in the number and variety of channels offered, as well as the included additional services, such as high-definition reception or digital video recorder services.
Bundled packages of the three categories (telephone, internet, and television) of residential telecommunications services are also included in each of the three indexes. Taken together, residential telecommunications services represent about 2.1 percent of the CPI as of December 2022. This percentage represents a slight decline from around 2.8 percent of the CPI 10 years ago, although the distribution of the expenditure on the services has moved toward television and internet services and away from residential telephone services. A factsheet is available on measuring price change of telecommunications services in the CPI.1
Pricing telecommunications services
The CPI program uses a cost-of-living framework to address questions that arise in constructing the CPI. However, the collection of prices for telecommunications services is complicated by challenges particular to these expenditure categories.
The challenges: initiation and substitution
Two main factors affected the accuracy of these residential telecommunications services: (1) initiation, selecting a specific eligible product or service from among those offered by the outlet surveyed for inclusion in the CPI sample, and (2) substitution, choosing an eligible product or service to replace one that is no longer for sale. Ordinarily, a new price quote is initiated into the CPI sample with the help of a respondent employed by the firm being surveyed. The respondent can provide information about the share of the firm’s revenue that is represented by a specific product offering. These revenue shares form the probability distribution used to select a unique item. That is, if an item represents 20 percent of a firm’s revenue, it should be selected for inclusion in the CPI sample 20 percent of the time.
A large proportion of sampled items for telecommunications services, however, is collected online rather than collected in person with a visit to a retail location. Hence, data are generally collected without the aid of a respondent. So, the usual process of selecting a unique eligible product to initiate into the CPI sample for pricing by using a probability based on the product’s proportion of the firm’s sales is not possible. This difficulty in initiating new items can lead to a reliance on equal probability between available plans in selecting a unique service plan to track over time. Without a respondent’s input about the share of the firm’s revenue each plan comprises, no mechanism exists to weight the various options.
Consumer expenditures are not likely equally distributed between various plans and options. Some plans are more popular with consumers than others and represent a disproportionate share of consumer expenditures. Commercial household survey data obtained by BLS show that more than 95 percent of consumers purchasing residential telephone services, for example, are purchasing bundled packages with one or more other telecommunications services.2 This propensity of consumers to purchase bundled services can introduce some mismeasurement into the index because standalone prices have risen, on average, more quickly than bundled prices, especially for telephone and internet services.
Telecommunications services indexes collected online are also subject to high rates of substitution as the offered plans change. Compared with other item categories in the CPI, telecommunications services plans in the CPI sample need to be substituted with other plans more often as they become unavailable. This increased rate of substitution is observed, in part, because the firms’ websites primarily provide information on new customer offerings, which can change rapidly. A review of commercial household survey data suggests that the CPI substitution rate for telecommunications services is 2 to 3 times larger than the actual rate at which consumers switch service plans.3 This finding reveals that many consumers will continue to purchase these services on contracts that are no longer being offered to new consumers and will not appear on the firms’ website. These substitutions are often to service plans that are not directly comparable to the plan previously priced for inclusion in the CPI. This result increases reliance on imputation to fill the uncollected observations with the average of quotes for similar items in the same area. Such increased reliance on imputation effectively reduces the sample size for these series, if quality adjustment techniques are not developed to allow such plans to be compared.
Addressing the challenges
To lessen the bias caused by relying on equal probability sampling in initiation, BLS began using commercial household survey data to guide field staff in selecting the most important characteristics and packages to price within new samples in February 2019. This new procedure allows CPI samples to include more bundled plans, mirror consumer behavior more closely, and reflect price changes more accurately.
To minimize the impact of substitutions between plans, CPI analysts developed five different hedonic regression models, one for each type of standalone plan (including wireless internet) and an additional model for bundled plans. These models help analysts adjust for the quality of substituted plans. The explanatory variables of interest with coefficients estimated by these models are the download speed and number of channels offered. Other categorical variables of interest for several calling features were also included in the model specifications, such as provided equipment, unlimited plans, premium packages, and various bundle combinations. The models also included a number of control variables, such as those for regional markets or specific firms. CPI analysts reestimate the models each year to ensure that the models remain relevant and continue to improve the accuracy of these series.
Conclusion
These methodological changes to the initiation and substitution processes for telecommunications services will improve the CPI estimates for these services. BLS is always working to improve the accuracy of the CPI series and will continue to introduce methodological improvements as appropriate.
United States Department of Labor
